PC peripheral power filtering cards.

[nonsuchpro]

Hi guys,
I just came across these on eBay and I,... sniff, sniFF, SNIFF? WOH! I smell something really bad.
Is it just my nose? Take a sniff and let me know.

PCI/PCI-E HiFi Power Filter card
http://www.ebay.com/itm/Elfidelity-PC-HI-FI-Power-Filter-card-PCI-PCI-E-HiFi-PC-audio-power-purification-/251894388743?talgo=origal&tfrom=251893473120&ttype=price&tpos=unknow

CPU/Memory HiFi DDR3 2V Power Supply Filter card
http://www.ebay.com/itm/Elfidelity-Power-purification-PC-HiFi-CPU-Memory-DDR3-2V-Power-Supply-Filter-USA-/251959406643?talgo=origal&tfrom=251893473120&ttype=price&tpos=unknow

SATA Power Noise Filter card
http://www.ebay.com/itm/Elfidelity-PC-Filtering-card-SATA-Power-Noise-Filter-purify-PC-HiFi-Hard-disk/251893473120?_trksid=p2047675.c100005.m1851&_trkparms=aid%3D222007%26algo%3DSIC.MBE%26ao%3D1%26asc%3D29979%26meid%3D0683995cdd4d4e399587fdb5078b2ed2%26pid%3D100005%26rk%3D1%26rkt%3D6%26sd%3D251894388743

[SeanB]

Shipped from the USA, Riiiiiiiiiiiiiiigggggggghhhhhhttt!!!!!!!!!!!!!!!

Bet the caps used are all the cheapest fakes they could pick up from the used cap market as well, pulled off dead electronics that went for recycling.

[NiHaoMike]

EVGA made something along the same lines:
http://www.evga.com/Products/Product.aspx?pn=100-MB-PB01-BR

[rs20]

So, I'm not going to buy this thing, but I've noticed that I can "hear" what my computer is doing through the headphones; evidently the DAC decoupling isn't up to scratch. Given the sheer bulk of capacitance on that PCI card, it actually seems reasonable to me that it could reduce audio frequency shenanigans happening on the power rails? I mean, not worth the money, I seriously doubt it'd do any more than cut the ripple* by a inaudible 10% or so... but it would sort of slightly work to reduce that ripple*, right?

* Where ripple is not the ripple corresponding to the DC-DC converters, but the lower-frequency power rail excursions due to changes in current draw.

[NiHaoMike]

One big contributor is that modern "low voltage" headphones, intended for use with mobile devices, need very little voltage to make a sound. Before resorting to more extreme measures, try making a simple adapter with 220 ohm or so resistors in series with each channel. (Try different values - go for the largest value that still gives enough volume.)

[amyk]

Not the first time this idea has been tried...
http://www.newegg.com/Product/Product.aspx?Item=N82E16814999012

...and they do have some effect, although it could be because the original decoupling was marginal already (anyone remember all the mobos with bad caps a few years ago?)
http://ixbtlabs.com/articles3/monitor/his-iclear.html

[German_EE]

Am I going to buy one of these? No, but I can see the reason behind their thinking. Put a scope on either the 3,3V or the 12V line in a PC and you can see all sorts of rubbish. I came across this issue years ago when I tried powering a pair of PC speakers off the 12V line in a PC and they were unusable because of all the background noise. So, any additional decoupling placed on a PC power supply line is a good thing.

However.........................

Good practice dictates that decoupling should be applied as close to the IC or device as possible. Those long PC traces will be inductive so I'm not sure how effective a decoupling card would be.

[rs20]

Those long PC traces will be inductive so I'm not sure how effective a decoupling card would be.

But we're only concerned about audio frequencies here, so the total reactance due to the impedance of those PC traces might be negligible.

[SeanB]

As most of the noise is going to be very fast fall time pulses, inductance of the traces and the connector will be very important. Remember the majority of the noise is at 100MHz or higher, with as well very strong 1-10MHz pulses from the on board CPU voltage regulator. 100MHz is the bus clocks, and pretty every chip will draw a pulse of current at that rate continuously with the clock running. The CPU voltage regulator will also have audible components in the ripple from the interleaving of the multiple converters ( ranges from 4-8 dseparate SMPS sections driven off a single controller) as well from the dynamic tracking of core voltage with compute load.

All in all a PC main board is a very noisy place, it easily will drown out radio near it even in a nominally shielded metal case.

[rs20]

As most of the noise is going to be very fast fall time pulses, inductance of the traces and the connector will be very important.

So we're assuming here that the sound circuitry is actually demodulating that high frequency noise somehow? Or downconverting through various nonlinearities? Because otherwise, the 100 MHz you noise you speak of will become 100 MHz noise in the headphones, i.e., inaudible.

[SeanB]

Trust me that opamps definitely have very poor PSSR at 100MHZ, and act as a pretty good demodulator of that noise........ The noise will be done by the internal nodes acting as slope rectifiers, mixing in the different noise sources and demodulating them into the audio band. Not much difference between a Gilbert mixer cell and your typical opamp, just that the opamp has some internal nodes that the Gilbert mixer uses as inputs connected to the supply rail. Noise on the supply will be downconverted.

[German_EE]

As most of the noise is going to be very fast fall time pulses, inductance of the traces and the connector will be very important.

So we're assuming here that the sound circuitry is actually demodulating that high frequency noise somehow? Or downconverting through various nonlinearities? Because otherwise, the 100 MHz you noise you speak of will become 100 MHz noise in the headphones, i.e., inaudible.

So, you have one interfering signal at 100.000 MHz and another signal at 100.001 MHz, both on the same power line. What do you get when you mix them together?

[rs20]

So, you have one interfering signal at 100.000 MHz and another signal at 100.001 MHz, both on the same power line. What do you get when you mix them together?

It wasn't clear to me that there were any mixers (i.e. nonlinear parts) around, and it's certainly not clear to me why a digital system would generate 100.001 MHz from a 100.000 MHz input. Nevertheless, I take SeanB's point.

[SeanB]

If you have spread spectrum turned on ( seems to be on by default on many BIOS models, as it reduces the apparent EMI in the EMI measurement by reducing the energy in a small band by spreading it all over) then you will get audio noise on the supply lines.  Just turned it off on my PC, and the audio output noise did reduce by a little, along with the zipper noise from the USB mouse.

[rollatorwieltje]

That linked soundcard is also nice.

http://www.ebay.com/itm/Elfidelity-Sound-Sourced-PCI-E-Hifi-AK4396-DAC-2-0-Channel-Internal-Sound-Cards-/271826440331?tfrom=251959406643&tpos=unknow&ttype=price&talgo=origal

It's probably a reasonable sound card, but look at those output opamps. Signetics 5534, a blast from the past!

I think I'll stick to my Xonar DX. The outputs are dead quiet on that card. My onboard sound is only useful as a random noise generator, it's horrible.

[senso]

Oh look, NEC-Tokin dreadfull tantaluns that killed thousands of Toshiba laptops, great choice, are they pulled dead from old board so they are already breaked-in(or broken?.. ).

[kingofkya]

So, I'm not going to buy this thing, but I've noticed that I can "hear" what my computer is doing through the headphones; evidently the DAC decoupling isn't up to scratch. Given the sheer bulk of capacitance on that PCI card, it actually seems reasonable to me that it could reduce audio frequency shenanigans happening on the power rails? I mean, not worth the money, I seriously doubt it'd do any more than cut the ripple* by a inaudible 10% or so... but it would sort of slightly work to reduce that ripple*, right?

* Where ripple is not the ripple corresponding to the DC-DC converters, but the lower-frequency power rail excursions due to changes in current draw.

Grab an asus audio card first card i have used that removes all of it. And there relatively cheap. And specifications for the noise floor.

https://www.asus.com/us/Sound_Cards/Xonar_DGX/

[NiHaoMike]

If you have spread spectrum turned on ( seems to be on by default on many BIOS models, as it reduces the apparent EMI in the EMI measurement by reducing the energy in a small band by spreading it all over) then you will get audio noise on the supply lines.  Just turned it off on my PC, and the audio output noise did reduce by a little, along with the zipper noise from the USB mouse.

My friend Tiffany Yep also hates the idea of spread spectrum clocking since it's very hard to predict and null out, especially if it's (pseudo) random modulated.

[amyk]

Oh look, NEC-Tokin dreadfull tantaluns that killed thousands of Toshiba laptops, great choice, are they pulled dead from old board so they are already breaked-in(or broken?.. ).

The problem with those Toshibas is they put the cap right under the CPU where it gets really hot, and it has only 1000h rating at 105C. It's not a tantalum, but polymer electrolytic.

[Zeyneb]

If you have spread spectrum turned on ( seems to be on by default on many BIOS models, as it reduces the apparent EMI in the EMI measurement by reducing the energy in a small band by spreading it all over) then you will get audio noise on the supply lines.  Just turned it off on my PC, and the audio output noise did reduce by a little, along with the zipper noise from the USB mouse.

Yeah the regulations for EMC are strict. And therefore companies implementing these things like clock modulation. But this will negatively affect the signal integrity for the product in order to operate correctly. So you would better turn this off when you enjoy seeing a result of 1 when 20 is divided by 20! Might help for other operations as well :-)

[Zeyneb]

Anyone having experience with the real luxurious American Technical Ceramics capacitors. Although expensive but they can provide the lowest amount of impedance. If you have them really close to the IC power pins. Trace impedance matters a lot as well.